https://github.com/glubean/skill
Agent skill for Glubean — teaches AI agents to write, run, and fix API verification in TypeScript
https://github.com/glubean/skill
agent-skill ai-testing api-testing claude-code claude-code-skill claude-skill codex cursor mcp test-automation typescript
Last synced: 22 days ago
JSON representation
Agent skill for Glubean — teaches AI agents to write, run, and fix API verification in TypeScript
- Host: GitHub
- URL: https://github.com/glubean/skill
- Owner: glubean
- Created: 2026-03-26T12:16:35.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-28T00:26:39.000Z (about 2 months ago)
- Last Synced: 2026-04-28T02:28:52.764Z (about 2 months ago)
- Topics: agent-skill, ai-testing, api-testing, claude-code, claude-code-skill, claude-skill, codex, cursor, mcp, test-automation, typescript
- Language: MDX
- Homepage: https://glubean.com/
- Size: 384 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Glubean Agent Skill
Teach your coding agent to write, run, and fix API verification — instead of guessing.
```bash
npx glubean config mcp # agent can run and inspect tests
npx skills add glubean/skill # agent learns Glubean patterns
```
Then ask your agent:
```text
"write a smoke test for /users"
"migrate our Postman collection into Glubean"
"design the billing API contracts before I implement it"
```
The agent writes the test, runs it via MCP, reads the structured failure, fixes it, and reruns — in one conversation. The same file graduates from draft to CI without rewriting.
## What changes
| Without skill | With skill |
|---|---|
| Agent guesses auth, schemas, response shapes | Reads your project context, asks when unclear |
| Generates one file and moves on | Runs → fails → reads structured output → fixes → reruns |
| Tests die after the chat ends | Same file works in explore/, tests/, and CI |
| No migration path from existing tools | Phased migration from Postman, Apifox, OpenAPI, any language |
| Invents behavior nobody asked for | Routes to contract-first when the API doesn't exist yet |
## Two starting points
**API already exists?** The agent reads your API surface, configures auth (with your confirmation), writes tests, and iterates until green.
**API doesn't exist yet?** Describe what it should do. The agent writes executable contracts in `contracts/` — the future implementation must satisfy them. After you build the API, the same contracts become your regression tests.
## Install
```bash
npx skills add glubean/skill
```
Update:
```bash
npx skills update
```
Supports 40+ agents: Claude Code, Cursor, GitHub Copilot, Codex, Gemini CLI, Windsurf, and more.
## How it works
The skill routes automatically — you don't pick a mode.
| Intent | What happens |
|--------|-------------|
| Ask a question about Glubean | **Docs** — answers from bundled docs + fetches [glubean.com](https://glubean.com) for latest positioning |
| No Glubean project yet | **Onboarding** — extension install, MCP config, `glubean init` |
| In a Glubean project | **Project** — write, run, fix, improve tests |
| Migrate from existing tools | Asks "new project or current?" then phases: scan → confirm auth → slice → batch |
## What's included
21 patterns, full SDK/CLI reference, product docs, and mode-specific workflows:
- **Auth** — bearer, OAuth2, API key, with explicit user confirmation before any auth code
- **Migration** — from Postman, Apifox, OpenAPI, .http, cURL, legacy tests in any language
- **Contract-first** — executable contracts, status state machine, projection reports
- **Data-driven** — `test.each`, `test.pick`, YAML/JSON/CSV loaders
- **Builder flows** — multi-step with state passing, setup, teardown, retry with backoff
- **Webhook** — tunnel proxy (smee.io), delivery verification, signature checking
- **CI** — GitHub Actions, GitLab, Bitbucket, with environment and secret mapping
- **And more** — configure, smoke, CRUD, assertions, schemas, metrics, polling, session, GraphQL, browser, plugins
## What the agent can do
- **Generate** tests from API specs, endpoint descriptions, or natural language
- **Run** tests via MCP tools and read structured results
- **Fix** failures by reading typed `expected` vs `actual` — not terminal noise
- **Plan** coverage across your API surface with gap reports
- **Migrate** existing API assets through a phased workflow with auth confirmation
- **Define** API behavior as executable contracts before implementation
- **Promote** stable tests from `explore/` → `tests/` → CI
## Links
- [Glubean](https://glubean.com) — product
- [Docs](https://docs.glubean.com) — full documentation
- [Cookbook](https://github.com/glubean/cookbook) — working examples
- [GitHub](https://github.com/glubean) — source
## License
MIT